Margin share between grower, processor, distributor and retailer varies more within a category than between categories. Contract structure — who carries input price risk and who owns the specification — predicts the split better than scale does.
Where does the margin actually sit?
Not where sector averages suggest. Variation within a category exceeds variation between categories, which means a benchmark drawn at category level tells you very little about a specific chain.
The determining variables are structural rather than positional. Two processors in the same category with different contract structures can sit at opposite ends of the margin distribution.
The averages are not wrong so much as useless at the level a decision is taken. A category-level margin split is a real average of real chains, and it describes none of them closely enough to underwrite one. It is a reasonable starting point for a question and a poor basis for an answer, which is the same relationship a cost curve has to a specific producer.
What moves margin along the chain?
Input price volatility and who absorbs it. When a contract passes input risk downstream, margin follows it upstream; when a retailer fixes price for a season, the processor becomes the shock absorber. Freight sits the same way, on contracts signed in a different quarter.
Because input volatility is episodic, margin position can look stable for years and then move sharply, which is why a static read of a chain is a poor basis for a multi-year thesis.
Who absorbs the volatility is also a matter of relative scarcity rather than contract language alone. A processor with a unique capability can push input risk back up the chain regardless of what the standard terms say, and one that is easily replaced cannot, whatever the contract nominally provides for. The written terms describe the position; substitutability decides whether it holds.
How do contract terms shift it?
Through risk allocation and specification ownership. Whoever specifies the product decides who can supply it, and that determines who is substitutable — which is what actually sets negotiating position.
A processor making to its own specification and a processor making to a retailer's specification are different businesses with the same equipment — and only one of them controls what happens at the next category reset.
Specification ownership also moves, which is what makes it worth asking about rather than assuming. A processor that has been making to its own recipe for a decade can find a retailer specifying the product at the next tender, and the margin consequence arrives with the specification rather than with any change in volume or cost.
Who can describe the real split?
Former commercial and procurement staff at each stage, and no single one of them sees the whole chain. That is why this question needs several conversations rather than one authoritative source, and why a panel across the stages resolves it faster than sequential calls.
Each participant knows their own margin and infers the others', usually incorrectly and usually in a direction that favors their own account.
Sequencing the conversations matters as much as the number of them. Starting at the retail end and working back tends to produce a chain where each participant's account is checked against the one downstream of it, which is where the inference errors show. Starting in the middle produces two unanchored halves.
What does this do to a thesis?
A margin-expansion thesis usually assumes the asset can hold price. If the contract passes input risk to it and it does not own the specification, that assumption is structural rather than operational.
Structural constraints do not respond to management action, which is the distinction that matters when a value creation plan is built on operational improvement.
Structural and operational constraints also fail differently under pressure. An operational shortfall shows up as underperformance against a plan; a structural one shows up as a plan that was never achievable, and the second is only visible in advance if somebody asked who owns the specification.